Position measurements of targets, in particular of structures on substrates in semiconductor manufacturing, are subject to various types of errors. Precise determination of the position of structures is important to assure that, ultimately, correctly functioning semiconductor products, like computer chips, for example, are produced. The demands with respect to precision increase as the structure dimensions on the chips to be produced decrease.
An important aspect of position measurements in the above context is registration of structures or sections of a structured surface with respect to each other. Errors of registration on a mask as determined by a measurement with a typical optical metrology tool, of which KLA-Tencor's LMS IPRO 5 is a contemporary example, may for example be due to errors in the optical metrology tool or to errors in the mask writer. By eliminating or reducing the errors occurring in the measurement with the optical metrology tool, the errors due to the mask writer can be identified.
For example, German Patent Application Publication No. DE 10 2008 060 293 A1 and United States Patent Application Publication No. US 2011/0229010 A1 disclose a method for determining relative positioning errors of plural sections of structures written on a substrate like a wafer or a photolithography mask. One magnified image of a region of the substrate larger than one section is recorded. Position errors of measurement marks contained in the image are determined from the image. The position errors are corrected for errors due to the imaging process. From the position errors corrected in this way the relative position error of the section is derived. This relative position error of a section is also known as stitching error, and the method assumes that errors due to the imaging process produce low frequency errors, whereas the stitching errors produce high frequency errors. Therefore, in order to remove the imaging errors, the low frequency error components are removed by a high-pass filtering process.
Another approach is to measure each target in an array of targets, for example each structure of interest in an arrangement of structures on an surface of a semiconductor substrate, individually, by moving the respective target into the center of the field of view of an imaging system of an optical metrology tool and performing the measurement.
The multi-region-of-interest registration measurement is a further approach. This makes use of the fact that often many targets are simultaneously contained in the field of view of an imaging system of an optical metrology tool. So the positions of plural targets, located at different positions relative to the field of view, can be measured at the same time.
However, the assumption made in the prior art about the mask writer having only high frequency errors is not strictly correct. By the high-pass filtering process information on the low frequency mask writer error therefore is discarded. In the case of individual target measurements, the throughput is very low. For example, on an IPRO4 metrology tool, measuring a single target may take up to 12 seconds, and measuring a typical array then up to 7 hours. During this long period of time, drift errors of the metrology tool can occur, which reduce the precision of the results.
In the multi-region-of-interest approach, due to optical distortion and aberrations which depend on the position in the field of view, different registration results may be produced if an array of targets like a mask with structures is shifted relative to the field of view and the position, relative to the array, or mask, respectively, coordinate system is determined for each shifted position. This error, depending on the field-of-view coordinates and also referred to as field-varying error, limits the achievable precision of registration measurements.
The optical error, like for example the optical distortion and/or aberrations, depends on the optical setup of the imaging system, but may also depend on parameters of the measured targets/structures, like size or symmetry of the targets or on proximity of two or more targets. The optical error can further depend on the substrate on which an array of targets is provided in specific technical fields, like in the case of wafers or masks in semiconductor manufacturing. There, the optical error can for example depend on the coatings, layer design or layer thickness of a mask.